sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling buggrafted

alb_break_llc() decides whether to break LLC preference during active
load balance. It does so by testing that every runnable fair task on the
source rq prefers its LLC:

	env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable

But the two counters cover different sets. nr_pref_llc_running is updated
in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued,
so it follows queued tasks. h_nr_runnable is updated in set_delayed()/
clear_delayed() and drops delay-dequeued tasks.

So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted
in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks,
alb_break_llc() returns false, and active balance is free to pull a task
off its preferred LLC. Active balance only moves runnable tasks, and this
is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB
skips the per-task test in can_migrate_task(). The runnable set is the one
we want.

Fix it on the counter side. A task should be counted in
nr_pref_llc_running exactly while it is both queued on its preferred LLC
(pref_llc_queued) and runnable (!sched_delayed). Define that membership
once in task_pref_llc_runnable(), and adjust the counter only through
pref_llc_running_inc()/pref_llc_running_dec() from the four sites that
change either input: account_llc_enqueue(), account_llc_dequeue(),
set_delayed() and clear_delayed(). Gating every update on the same
predicate keeps the delay, wake and dequeue paths from double-counting
or underflowing; see the comments at those sites for the ordering.

nr_llc_running and sd->llc_counts are not touched and stay on queued
semantics.

Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC")
Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/
Reported-by: Zhan Xusheng <zhanxusheng@xiaomi.com>
Suggested-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Kayra Cizmeci <kayracizmeci@gmail.com>
Cc: <stable@kernel.org> # v7.2.x
Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com

1 files changed, 100 insertions, 0 deletions

diff --git a/fs/xfs/xfs_sysctl.h b/fs/xfs/xfs_sysctl.h
new file mode 100644
index 000000000..ed9d89607
--- /dev/null
+++ b/fs/xfs/xfs_sysctl.h
@@ -0,0 +1,100 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2001-2005 Silicon Graphics, Inc.
+ * All Rights Reserved.
+ */
+#ifndef __XFS_SYSCTL_H__
+#define __XFS_SYSCTL_H__
+
+#include <linux/sysctl.h>
+
+/*
+ * Tunable xfs parameters
+ */
+
+typedef struct xfs_sysctl_val {
+	int min;
+	int val;
+	int max;
+} xfs_sysctl_val_t;
+
+typedef struct xfs_param {
+	xfs_sysctl_val_t panic_mask;	/* bitmask to cause panic on errors. */
+	xfs_sysctl_val_t error_level;	/* Degree of reporting for problems  */
+	xfs_sysctl_val_t syncd_timer;	/* Interval between xfssyncd wakeups */
+	xfs_sysctl_val_t stats_clear;	/* Reset all XFS statistics to zero. */
+	xfs_sysctl_val_t inherit_sync;	/* Inherit the "sync" inode flag. */
+	xfs_sysctl_val_t inherit_nodump;/* Inherit the "nodump" inode flag. */
+	xfs_sysctl_val_t inherit_noatim;/* Inherit the "noatime" inode flag. */
+	xfs_sysctl_val_t inherit_nosym;	/* Inherit the "nosymlinks" flag. */
+	xfs_sysctl_val_t rotorstep;	/* inode32 AG rotoring control knob */
+	xfs_sysctl_val_t inherit_nodfrg;/* Inherit the "nodefrag" inode flag. */
+	xfs_sysctl_val_t fstrm_timer;	/* Filestream dir-AG assoc'n timeout. */
+	xfs_sysctl_val_t blockgc_timer;	/* Interval between blockgc scans */
+} xfs_param_t;
+
+/*
+ * xfs_error_level:
+ *
+ * How much error reporting will be done when internal problems are
+ * encountered.  These problems normally return an EFSCORRUPTED to their
+ * caller, with no other information reported.
+ *
+ * 0	No error reports
+ * 1	Report EFSCORRUPTED errors that will cause a filesystem shutdown
+ * 5	Report all EFSCORRUPTED errors (all of the above errors, plus any
+ *	additional errors that are known to not cause shutdowns)
+ *
+ * xfs_panic_mask bit 0x8 turns the error reports into panics
+ */
+
+enum {
+	/* XFS_REFCACHE_SIZE = 1 */
+	/* XFS_REFCACHE_PURGE = 2 */
+	/* XFS_RESTRICT_CHOWN = 3 */
+	XFS_SGID_INHERIT = 4,
+	XFS_SYMLINK_MODE = 5,
+	XFS_PANIC_MASK = 6,
+	XFS_ERRLEVEL = 7,
+	XFS_SYNCD_TIMER = 8,
+	/* XFS_PROBE_DMAPI = 9 */
+	/* XFS_PROBE_IOOPS = 10 */
+	/* XFS_PROBE_QUOTA = 11 */
+	XFS_STATS_CLEAR = 12,
+	XFS_INHERIT_SYNC = 13,
+	XFS_INHERIT_NODUMP = 14,
+	XFS_INHERIT_NOATIME = 15,
+	XFS_BUF_TIMER = 16,
+	XFS_BUF_AGE = 17,
+	/* XFS_IO_BYPASS = 18 */
+	XFS_INHERIT_NOSYM = 19,
+	XFS_ROTORSTEP = 20,
+	XFS_INHERIT_NODFRG = 21,
+	XFS_FILESTREAM_TIMER = 22,
+};
+
+extern xfs_param_t	xfs_params;
+
+struct xfs_globals {
+#ifdef DEBUG
+	int	pwork_threads;		/* parallel workqueue threads */
+	bool	larp;			/* log attribute replay */
+#endif
+	int	bload_leaf_slack;	/* btree bulk load leaf slack */
+	int	bload_node_slack;	/* btree bulk load node slack */
+	int	log_recovery_delay;	/* log recovery delay (secs) */
+	int	mount_delay;		/* mount setup delay (secs) */
+	bool	bug_on_assert;		/* BUG() the kernel on assert failure */
+	bool	always_cow;		/* use COW fork for all overwrites */
+};
+extern struct xfs_globals	xfs_globals;
+
+#ifdef CONFIG_SYSCTL
+extern int xfs_sysctl_register(void);
+extern void xfs_sysctl_unregister(void);
+#else
+# define xfs_sysctl_register()		(0)
+# define xfs_sysctl_unregister()	do { } while (0)
+#endif /* CONFIG_SYSCTL */
+
+#endif /* __XFS_SYSCTL_H__ */